Well, this is a bug, for a way around, please insert an additional space inside <!DOCTYPE to make it <! DOCTYPE.
We have a file content recognizer that tries to estimate content type without parsing xml, and in this case it does not take into account that doctype declaration is commented out.